How Quanto-Adjustment Algorithms Price Foreign Assets in a Fixed Currency: Numeraire Changes, Covariance Drift, Equity–FX Correlation, Greeks and Model Risk

Reader question: If an option depends on a foreign asset but pays in a fixed domestic-currency conversion rate, why does foreign-exchange volatility still affect the price even though the final FX rate is not used in the payoff?

A quanto removes the realised exchange rate from the payoff, but it does not remove the economic dependence between the foreign asset and the exchange rate. When we switch from the foreign pricing measure to the domestic pricing measure, the asset’s drift changes by a covariance term. That quanto adjustment contains the product of asset volatility, FX volatility and their correlation.

This article owns the constant-parameter quanto drift-adjustment problem: foreign asset + domestic/foreign rates + asset volatility + FX volatility + correlation + fixed conversion payoff → domestic-measure drift, adjusted forward, option price, correlation sensitivity and diagnostics.

It does not own general cross-currency basis modelling, stochastic-correlation models, local/stochastic volatility calibration or personalised hedging. Those are separate roles.

This is public mathematical and computational education. It is not financial advice or a recommendation to trade cross-currency derivatives.

1. Define the currency convention before doing any mathematics

Sign mistakes in quanto pricing often begin with an unstated FX convention.

In this article define:

X = domestic currency units per one unit of foreign currency.

For example, if domestic currency is SGD and foreign currency is USD, X could be SGD per USD.

Let:

• S = foreign asset price measured in foreign currency;
• r_d = domestic risk-free rate;
• r_f = foreign risk-free rate;
• q = foreign asset dividend yield;
• σ_S = asset volatility;
• σ_X = FX volatility;
• ρ = instantaneous correlation between asset and FX returns.

If the FX quote is inverted, the sign of the correlation term changes. A production engine should store the quote convention as data rather than assume it silently.

2. What makes a payoff a quanto?

A simple foreign-equity quanto call can pay at maturity:

Q · max(S_T − K, 0)

in domestic currency, where Q is a conversion rate fixed in advance.

The actual maturity FX rate X_T is absent from the payoff.

Contrast this with a “compo” or ordinary converted payoff:

X_T · max(S_T − K, 0).

The two claims have different risks. The quanto fixes the conversion rate but embeds correlation risk through the pricing measure.

3. Under the domestic measure, the FX drift is fixed by no-arbitrage

With the convention X = D/F, a constant-rate lognormal FX model under the domestic risk-neutral measure has:

dX/X = (r_d − r_f)dt + σ_XdW_X^d.

This is the continuous-time version of covered-interest-parity logic.

The foreign asset measured in foreign currency would naturally have drift r_f−q under the foreign risk-neutral measure. But a quanto payoff is discounted and valued in domestic currency, so we need the foreign asset dynamics under the domestic measure.

4. The domestic-measure drift contains the quanto adjustment

Let the Brownian shocks satisfy:

dW_S^ddW_X^d = ρdt.

Under the domestic measure, the foreign asset dynamics become:

dS/S = [r_f − q − ρσ_Sσ_X]dt + σ_SdW_S^d.

The covariance term:

−ρσ_Sσ_X

is the basic quanto drift adjustment for this quote convention.

If ρ = 0, there is no adjustment. If either volatility is zero, there is no adjustment. The term exists only because two stochastic quantities covary.

5. A no-arbitrage derivation from the domestic value XS

The domestic-currency value of the foreign asset is:

Z = XS.

Using Itô’s product rule:

dZ/Z = dX/X + dS/S + (dX/X)(dS/S).

The covariance term contributes:

ρσ_Xσ_Sdt.

A tradable foreign equity converted into domestic currency must have the appropriate domestic risk-neutral drift after accounting for its dividend yield. Substituting the FX drift and solving for the required drift of S produces:

μ_S^d = r_f − q − ρσ_Sσ_X.

This provides a useful implementation invariant. If a coded quanto drift does not make the domestic converted asset obey the correct martingale relationship, the sign convention is wrong.

6. The adjusted quanto forward

Under constant parameters, the domestic-measure expectation of S_T is:

F_q = S₀ exp[(r_f − q − ρσ_Sσ_X)T].

This is the quanto-adjusted forward of the foreign asset for the fixed-currency payoff.

It differs from the plain foreign forward:

F_f = S₀ exp[(r_f − q)T]

by:

F_q/F_f = exp[−ρσ_Sσ_XT].

7. Correlation direction under the chosen convention

With X = domestic per foreign:

• ρ > 0: the quanto drift is reduced;
• ρ = 0: the foreign forward is unchanged;
• ρ < 0: the quanto drift is increased.

For a standard call with all else equal, more positive correlation therefore tends to reduce the quanto-adjusted forward and call value under this convention.

If another source defines FX as foreign per domestic, the correlation sign must be transformed consistently. Comparing formulas without comparing quote conventions is a common source of apparent disagreement.

8. Closed-form quanto call under constant volatilities

Because S remains lognormal under the domestic measure in this simple model, a Black-style formula applies using the adjusted forward.

For fixed conversion rate Q:

C_q = Qe^−r_dT[F_qN(d₁) − KN(d₂)],

where:

d₁ = [ln(F_q/K) + ½σ_S²T]/(σ_S√T),

d₂ = d₁ − σ_S√T.

A put follows the corresponding Black put formula.

Notice that σ_X does not appear inside the option’s terminal conditional variance in this simple quanto formula; it enters through the adjusted forward via covariance. More advanced models can change that separation.

9. A numerical illustration

Suppose:

• S₀ = 100;
• r_f − q = 2%;
• σ_S = 20%;
• σ_X = 10%;
• ρ = 0.50;
• T = 1 year.

The covariance adjustment is:

ρσ_Sσ_X = 0.50 × 0.20 × 0.10 = 1%.

The domestic-measure drift of the foreign asset is therefore:

2% − 1% = 1%.

The plain foreign forward uses 2%, while the quanto-adjusted forward uses 1%.

The example is deliberately simple. The point is not the exact option premium; it is that correlation changes a drift under the pricing measure even though the realised FX rate is absent from the final payoff.

10. Why this is a change-of-numeraire problem

Risk-neutral drifts are not physical forecasts. They depend on the numeraire and associated pricing measure.

Under the foreign money-market numeraire, the foreign equity’s ex-dividend drift is naturally tied to r_f−q. Under the domestic money-market numeraire, the Radon–Nikodym change of measure shifts Brownian drifts according to covariance with the numeraire ratio.

The quanto adjustment is therefore a concrete application of a broader principle:

change the numeraire → change the measure → change the drift.

11. Correlation is an input, not a decorative statistic

The price sensitivity to correlation can be meaningful because correlation changes the entire adjusted forward exponentially with maturity.

A simple derivative of the log quanto forward is:

∂ ln F_q/∂ρ = −σ_Sσ_XT.

This makes the direction transparent.

The actual option-price sensitivity also depends on delta with respect to the adjusted forward.

12. Correlation risk grows with horizon and volatilities

The magnitude of the simple adjustment scales with:

|ρ|σ_Sσ_XT.

Therefore the constant-parameter formula predicts stronger quanto effects when:

• asset volatility is high;
• FX volatility is high;
• absolute correlation is large;
• maturity is long.

That does not mean the constant-correlation model becomes more reliable in those regimes. In fact, model risk can also increase because correlation and volatility surfaces are less likely to remain constant.

13. Inputs and outputs

Inputs can include:

• foreign asset spot;
• strike and maturity;
• fixed conversion rate;
• domestic discount curve;
• foreign rate/dividend curve;
• asset volatility;
• FX volatility;
• asset–FX correlation;
• FX quote convention;
• payoff currency;
• settlement convention.

Outputs can include:

• quanto drift adjustment;
• quanto-adjusted forward;
• option price;
• delta and vega;
• correlation sensitivity;
• FX-volatility sensitivity through the adjustment;
• scenario table across ρ;
• analytic-versus-Monte-Carlo reconciliation.

14. A diagnostic triangle: three models should agree in the simple case

For constant rates and volatilities, three independent routes can be compared:

1. the closed-form quanto formula;
2. a two-factor Monte Carlo simulation under the domestic measure;
3. a two-dimensional PDE with the correct mixed derivative and payoff.

Agreement across three different numerical representations is stronger evidence than agreement within one code path.

15. Monte Carlo benchmark

Simulate correlated Brownian increments:

Z_X = ρZ_S + √(1−ρ²)Z_⊥.

Under the domestic measure, simulate S using the adjusted drift and discount the fixed-currency payoff with the domestic discount factor.

The Monte Carlo mean should converge to the closed-form result within its sampling error.

Falsifier: if increasing paths shrinks Monte Carlo error bars around a value different from the analytic price, the measure, drift, correlation or payoff convention is inconsistent.

16. Zero-correlation and zero-volatility invariants

Several edge cases should be exact.

• If ρ = 0, the quanto adjustment is zero.
• If σ_X = 0, the adjustment is zero.
• If σ_S = 0, the covariance adjustment is zero and the payoff becomes deterministic conditional on rates.

These are useful software unit tests because they isolate the covariance machinery.

17. Evidence polarity

Evidence for confidence includes exact edge-case behaviour, domestic-value martingale consistency, analytic/Monte Carlo agreement, stable results under correlation stress, calibrated volatilities from coherent asset and FX surfaces, explicit quote conventions and small residuals when the model is compared with liquid quanto instruments.

Evidence against confidence includes a correlation sign that changes when the same FX quote is merely relabelled, different departments using opposite FX conventions, large unexplained discrepancies between analytic and simulation engines, correlation estimates unstable across windows, flat-volatility assumptions applied across strong smiles, or prices highly sensitive to extrapolated correlations with little market evidence.

18. Counterexample: wrong FX quote convention

Suppose one implementation uses SGD/USD while another uses USD/SGD but both feed the same numerical ρ into the formula.

Inverting an FX rate reverses the sign of its infinitesimal return to first order, so the corresponding correlation changes sign.

Falsifier: invert the FX quote and transform the correlation consistently. The economic price should remain invariant after all conventions are transformed. If it does not, the implementation contains a convention bug.

19. Counterexample: constant correlation hides state dependence

The simple quanto formula assumes one constant ρ. Real asset–FX dependence can vary across time, stress regimes and return direction.

Stochastic-correlation and copula models can produce materially different prices when dependence is asymmetric or time varying.

Falsifier: estimate correlation across rolling windows and stress regimes. If the input is unstable and price sensitivity to ρ is large, a single historical correlation should not be presented as a precise truth.

20. Counterexample: smiles make constant volatility insufficient

A foreign equity may have a strong volatility smile, and the FX market may have its own smile. A constant σ_S and σ_X reduce each surface to one number.

Even if ATM vanilla prices are matched, a quanto with nonlinear exposure can remain mispriced.

Falsifier: compare model prices across strikes and maturities rather than only at one ATM point.

21. Counterexample: stochastic rates and cross-currency basis

The closed-form derivation above assumes simple deterministic rates. Long-dated cross-currency products can depend materially on stochastic rates, collateral currency and cross-currency basis.

Those effects alter discounting and measure changes. A single textbook adjustment cannot automatically absorb them.

Falsifier: perturb domestic and foreign curve dynamics or basis assumptions. If the product’s value changes materially, a richer joint model is required.

22. Counterexample: jumps and tail dependence

Joint equity–FX moves during market stress may contain jumps or tail dependence that a correlated Gaussian Brownian model cannot represent.

A model can match ordinary correlation while underrepresenting extreme co-movement.

Falsifier: compare joint-tail observations and stress scenarios with the Gaussian copula implied by the constant-correlation model.

23. Correlation from history versus correlation from prices

Historical return correlation is a physical-measure statistic. A pricing model is risk-neutral and may embed a correlation risk premium.

If liquid quanto or related cross-asset options exist, their prices can contain information about risk-neutral dependence that differs from historical estimates.

A robust process should distinguish:

• historical correlation for descriptive risk;
• risk-neutral/implied correlation for pricing;
• stress correlations for risk management.

24. Alternatives and extensions

Stochastic-correlation models allow ρ itself to evolve.

Local or stochastic volatility models fit asset and FX smiles more flexibly.

Copula models separate marginal distributions from dependence structure.

Joint Heston/Wishart-type models can make volatilities and covariances stochastic.

Monte Carlo handles richer multi-factor dynamics but adds simulation and calibration cost.

PDE methods can solve low-dimensional joint models with explicit state dependence.

25. Connections to the surrounding Bukit Timah Tutor knowledge estate

The measure-change foundation is Girsanov change-of-measure algorithms. Quanto adjustment is a concrete cross-currency application of changing pricing measures.

The FX no-arbitrage baseline is FX-forward pricing algorithms.

More complicated funding interactions connect to cross-currency swap and FX-swap algorithms.

The option formula itself inherits the vanilla logic explained in Black–Scholes option-pricing algorithms.

For richer volatility dynamics, see Heston calibration algorithms.

The full lane is indexed at Finance & Banking Algorithms | Applied Mathematics in Real Financial Systems.

26. What would falsify confidence?

Confidence should be withdrawn if the domestic converted asset fails its martingale/no-arbitrage invariant; if analytic and Monte Carlo engines disagree beyond sampling error; if quote inversion changes the economic price; if correlation and volatility inputs are unstable or unsupported; if smile, basis, stochastic-rate or tail effects are material to the use case; or if a physical historical correlation is silently treated as a risk-neutral pricing parameter.

27. Verification and update triggers

Preserve the domestic/foreign currency labels, FX quote orientation, curves, asset and FX volatility inputs, correlation source, fixed conversion rate, payoff definition and pricing measure with every valuation.

Revalidate after changes to FX quote conventions, collateral currency, cross-currency basis, asset or FX volatility surfaces, correlation calibration, stochastic-rate assumptions or payoff conversion rules. Trigger review when realised or implied dependence shifts materially, especially during regime changes when the constant-correlation approximation is most vulnerable.

28. Primary and high-quality references

• Hélyette Geman, Nicole El Karoui and Jean-Charles Rochet, Changes of numéraire, changes of probability measure and option pricing, Journal of Applied Probability, 1995.
• Tomas Björk, Change of Numeraire, Encyclopedia of Quantitative Finance, 2010.
• Nicole Branger and Matthias Muck, Keep on Smiling? Volatility Surfaces and the Pricing of Quanto Options when all Covariances are Stochastic, Journal of Banking & Finance, 2012.
• Modelling stochastic correlation, Journal of Mathematics in Industry, including a quanto-pricing application.
• Rafael Prudencio and Christian Jäkel, The Pricing of Quanto Options: An empirical copula approach, for a dependence-model alternative.

Educational boundary: The quanto adjustment is a no-arbitrage covariance correction under a stated measure, quote convention and model. It is not a forecast of future FX or equity returns.